RAC1P29S Induces a Mesenchymal Phenotypic Switch via Serum Response Factor to Promote Melanoma Development and Therapy Resistance

Abstract

RAC1 P29 is the third most commonly mutated codon in human cutaneous melanoma, after BRAF V600 and NRAS Q61. Here, we study the role of RAC1^P29S in melanoma development and reveal that RAC1^P29S activates PAK, AKT, and a gene expression program initiated by the SRF/MRTF transcriptional pathway, which results in a melanocytic to mesenchymal phenotypic switch. Mice with ubiquitous expression of RAC1^P29S from the endogenous locus develop lymphoma. When expressed only in melanocytes, RAC1^P29S cooperates with oncogenic BRAF or with NF1-loss to promote tumorigenesis. RAC1^P29S also drives resistance to BRAF inhibitors, which is reversed by SRF/MRTF inhibitors. These findings establish RAC1^P29S as a promoter of melanoma initiation and mediator of therapy resistance, while identifying SRF/MRTF as a potential therapeutic target.

Keywords: BRAF; EMT; MRTF; NF1; PTEN; RAC1; SRF; melanoma; p53.

Graphical abstract

Figure 1

Effect of Activation of RAC1^P29S on Survival of Melanocytes (A) Schematic of the ER-RAC1^P29S fusion protein system. ER, estrogen receptor; 4OHT, 4-hydroxytamoxifen; HSP, heat shock protein. (B) RAC1-GTP assay in MCF10A cells (left) and mouse melanocyte cell line melan-a (right) stably expressing ER-RAC1^P29S. Cells were treated with 500 nM 4OHT for 24 h and ER-RAC1^P29S binding to the PAK1 RAC1 binding domain was assessed using a pull-down assay. (C) PAK1/2 and ERK1/2 phosphorylation by ER-RAC1^P29S in melanocytes. Cells were treated with 500 nM 4OHT for 24 h. (D) Quantification of phospho-PAK1/2 and phospho-ERK1/2 levels detected by immunoblotting, normalized using vinculin loading control (n = 3 or more independent experiments); t test versus 0 h was used with Holm-Sidak correction for multiple testing; ^∗p < 0.05, ^∗∗∗p < 0.001; n.s., not statistically significant. (E) Morphological changes induced by ER-RAC1^P29S activation in melan-a melanocytes treated with 500 nM 4OHT. (F) Viability of melan-a melanocytes with ER-RAC1^P29S in growth factor-reduced medium (fetal calf serum [FCS] 0.25%, 12-O-tetradecanoylphorbol-13-acetate [TPA]-free). Cells were treated with 1 μM 4OHT and viability was quantified using CellTiter-Glo (n = 3 independent experiments); t test was used for statistical comparison. (G) Soft agar sphere formation of melan-a melanocytes with activated ER-RAC1^P29S. Cells were grown in full medium containing 1 μM 4OHT for 3 weeks before staining, imaging, and automated counting (n = 3 biological replicates). (H) Effect of activation of ER-RAC1^P29S on apoptosis in melan-a melanocytes. Cells were treated with 1 μM 4OHT. Cleaved caspase-3 was quantified and normalized to cell viability (n = 2 independent experiments, three biological replicates per experiment). Bars represent means ± SD. See also Figure S1.

Figure 2

Effect of Acute Activation of RAC1^P29S in Melanocytes on AKT and SRF/MRTF Effector Pathways (A) Reverse-phase protein array using ER-RAC1^P29S melanocytes after RAC activation, expressed as a scatterplot with p values and log2 fold change compared with 0 h control (means from 3 independent experiments). (B) Densitometric quantification of phospho-AKT levels determined by immunoblotting, normalized to total AKT levels. Melan-a ER-RAC1^P29S cells were treated with 4OHT at 500 nM (n = 4 independent experiments, one representative immunoblot shown); t test was used for statistical comparison. (C) RNA sequencing (RNA-seq) in ER-RAC1^P29S melanocytes, expressed as scatterplots for empty vector cells and ER-RAC1^P29S cells treated with 500 nM 4OHT. Blue dots represent values statistically different to 0 h values (false discovery rate [FDR] 5%). (D) Transcription factor target enrichment analysis of gene expression changes in melanocytes 4 h after activation of ER-RAC1^P29S. Gene set enrichment analysis (GSEA) using the transcription factor target gene sets from the MSigDB database (Broad). NES, normalized enrichment score. (E) GSEA plot of the direct SRF target gene set from Esnault et al. (2014), using all genes expressed in the ER-RAC1^P29S melanocytes at 4 versus 0 h 4OHT treatment. (F) Normalized RNA-seq read counts of canonical SRF/MRTF targets (n = 3 independent experiments). (G) Pathway enrichment analysis of gene expression changes in melanocytes 40 h after activation of ER-RAC1^P29S. GSEA using the curated hallmark gene set collection from MSigDB. (H) GSEA plot of the epithelial-to-mesenchymal transition (EMT) hallmark gene set from MSigDB using gene expression differences in ER-RAC1^P29S melanocytes treated for 40 h with 4OHT compared with 0 h control. (I) Relative mRNA of mesenchymal markers in melanocytes with activated ER-RAC1^P29S, quantified using RNA-seq (n = 3 independent experiments). (J) Effect of treatment with SRF/MRTF inhibitors on mesenchymal marker induction by ER-RAC1^P29S. Cells were pre-treated with 5–10 μM CCG-1423 or CCG-203971 1 h before 4OHT treatment (1 μM) for 48 h (n = 3 independent experiments); mRNA levels were determined using qPCR and normalized to Gapdh expression. For all graphs: bars represent means ± SD. See also Figure S2.

Figure 3

Impact of Endogenous RAC1^P29S on AKT and SRF/MRTF Activation in Melanocytes and Mesenchymal Phenotype (A) Schematic of the Rac1^LSL−P29S allele. (B) Rosa26-CreER^+/–;Rac1^{LSL−P29S/WT} and Rosa26-CreER^+/–;Rac1^WT/WT MEFs were isolated and recombined using 1 μM 4OHT. Activation of RAC1 was assessed using pull-down assay. (C) Phosphorylation of PAK1/2 and AKT in MEFs upon recombination of the Rac1^LSL−P29S allele by 4OHT treatment (1 μM). (D) Phosphorylation of PAK1/2 and AKT in mouse melanocytes with endogenous RAC1^P29S. Immunoblot of three independent cultures per genotype is shown. (E) Transcription factor target enrichment analysis of gene expression changes in melanocytes with endogenous RAC1^P29S versus RAC1^WT. GSEA was performed using the transcription factor target gene set collection from MSigDB in combination with RNA-seq data of melanocytes from Rac1^{LSL−P29S/WT} mice versus Rac1^WT/WT mice (n = 3 independent cultures per genotype). (F) GSEA results using SRF/MRTF/TCF gene sets from Esnault et al. (2014), was applied on RNA-seq data from melanocytes cultured with endogenous RAC1^P29S versus RAC1^WT. GSEA plot for direct targets of MRTF is shown. (G) Normalized mRNA read counts of canonical SRF/MRTF targets in melanocytes with endogenous RAC1^P29S (n = 3 independent cultures per genotype); Wald test was applied in combination with Benjamini-Hochberg correction. (H) GSEA plot of the curated EMT hallmark gene set from MSigDB using RNA-seq data of melanocytes from Rac1^{LSL−P29S/WT} mice versus Rac1^WT/WT mice (n = 3 independent cultures per genotype). (I) Effect of SRF/MRTF pathway inhibition on mRNA expression of vimentin in mouse melanocytes with endogenous RAC1^P29S. Cells were treated with 10–25 μM CCG-1423 or CCG-203971 and starved of TPA for 24 h (n = 3 independent experiments). For all graphs: bars represent means ± SD. See also Figure S3.

Figure 4

Discovery of AKT and MRTF Dependencies in Cells with Endogenous RAC1^P29S (A) Melanocyte cultures from Rac1^WT/WT and Rac1^{LSL−P29S/WT} mice were grown in growth factor-reduced medium (FCS 0.25%, TPA-free) and assayed for viability using CellTiter-Glo; p value using two-way ANOVA for the genotype indicated. (B) Single-cell suspensions of melanocytes from Rac1^WT/WT and Rac1^{LSL−P29S/WT} mice (n = 3 independent cultures per genotype) were seeded in soft agar for 3 weeks before imaging and automated counting. (C) Melanocytes from Rac1^WT/WT and Rac1^{LSL−P29S/WT} mice were transfected with a custom siRNA library (n = 205 pools), grown in TPA-free medium for 3 days and assayed for viability (n = 3 independent cultures per genotype, each assayed in three independent experiments). (D) Detailed plots of the data from the RAC1 effector siRNA screen presented in (C). Control siRNAs (top) and siRNAs that specifically kill melanocytes with endogenous RAC1^P29S (bottom). (E) Effect of combination treatments with inhibitors of the SRF/MRTF, PAK, and AKT pathways on viability of melanocytes with RAC1^WT (left) or endogenous RAC1^P29S (right). Mouse melanocyte cultures were grown in regular medium and assayed for viability 72 h after starting drug treatments. Inhibitor doses: CCG-203971 (SRF/MRTF) 10 μM, G-5555 (PAK) and MK-2206 (AKT) 2 μM. (F) Effects on viability of human melanoma cell lines of reduction in expression of RAC1 or BRAF. Cells were cultured in growth medium and assayed for viability at 96 h post-transfection (n = 5 cell lines per genotype for the BRAF panel, n = 7 cell lines per genotype for the RAC1 panel, dots represent means of a single cell line that was tested in multiple independent experiments); p values from t test. (G) Drug panel targeting RAC1 effectors applied in human melanoma cell lines with RAC1^P29S versus cell lines with RAC1^WT. Cells were cultured in growth medium and assayed for viability 72 h after starting drug treatments. Drugs ranked by the percentage difference between the half maximal inhibitory concentration (μM) of RAC1^WT cell lines and RAC1^P29S cell lines. For each drug, two-way ANOVA was used to probe the full dataset for a genotype effect, which was corrected according to Benjamini-Hochberg to produce FDR q values; N.D., not determined because curve fitting failed. (H) Viability of human melanoma cell lines treated with the SRF/MRTF inhibitor CCG-1423 was determined using CellTiter-Glo (n = 7 cell lines per genotype, for each cell line the mean from two independent experiments was used); statistical analysis was performed as described in (F). (I) Apoptosis of mouse melanocytes with endogenous RAC1^P29S or RAC1^WT cultured in growth medium and treated with obatoclax at 250 nM, 500 nM, or 1 μM for 3 days (n = 2–3 independent cultures per genotype). (J) Schematic of the SRF/MRTF signaling pathway. For all graphs: bars represent means ± SD. See also Figure S4.

Figure 5

Effect of Ubiquitous Expression of RAC1^P29SIn Vivo on Induction of B Cell Lymphoma (A) Schematic of PGK-Cre/Rac1^LSL−P29S cross and summary of results from the timed analysis of embryos. PGK-Cre was always maternal to ensure germline Cre recombination in all embryos. p values derived from chi-square testing are indicated. (B) Representative micrographs showing gross cardiovascular abnormalities and developmental retardation in Rac1^{LSL−P29S/WT} embryos (E10.5). Red arrow, enlarged pericardial cavity. (C) Schematic of experimental design to express RAC1^P29S in the whole body of adult mice. Tamoxifen was administered by oral gavage. (D) Representative image of the spleen and quantification of spleen weight in aged mice with the indicated genotype. Bars represent means ± SD (n = 8–13 mice per genotype); Mann-Whitney test used for statistical comparison. (E) Tumor-free survival curve of mice with indicated genotypes after treatment with tamoxifen. Result from log rank testing (Mantel-Cox) is indicated. (F) Representative photos of the digestive tract from a Rosa26-CreER^+/–;Rac1^WT/WT mouse and a mesenteric lymphoma-bearing Rosa26-CreER^+/–;Rac1^{LSL−P29S/WT} mouse. Red arrow, lymphoma. (G) Representative H&E- and B220-stained sections of a mesenteric lymphoma versus normal mesenteric lymph nodes in a control mouse. (H) Representative H&E- and B220-stained sections of a normal spleen from a Rosa26-CreER^+/–;Rac1^WT/WT mouse and the spleen, thymic lymphoma, and squamous cell tumor of the skin from Rosa26-CreER^+/–;Rac1^{LSL−P29S/WT} mice. See also Figure S5.

Figure 6

Effect of Melanocytic Expression of RAC1^P29SIn Vivo on Tumorigenesis (A) Schematic of the Tyr-CreER, Braf^CA and Rac1^LSL−P29S allele combination. (B) Melanoma-free survival curves of Tyr-CreER^+/–;Braf^CA/WT;Rac1^WT/WT mice and Tyr-CreER^+/–;Braf^CA/WT;Rac1^{LSL−P29S/WT} mice were compared using log rank testing (Mantel-Cox). (C) Effects of RAC1^P29S on tumor number in Tyr-CreER^+/–;Braf^CA/WT mice. Groups were compared using the Mann-Whitney test (n = 12–13 mice per genotype). (D) Schematic of the Tyr-CreER, Pten^F, Braf^CA, and Rac1^LSL-P29S allele combination. (E) Overall survival curves of Tyr-CreER^+/–;Pten^F/WT;Braf^CA/WT;Rac1^WT/WT mice and Tyr-CreER^+/–;Pten^F/WT;Braf^CA/WT;Rac1^{LSL−P29S/WT} mice were compared using the log rank testing (Mantel-Cox). (F) Effects of RAC1^P29S on tumor number in Tyr-CreER^+/–;Pten^F/WT;Braf^CA/WT mice. Groups compared using Mann-Whitney testing (n = 5–9 mice per genotype). (G) Schematic of the Tyr-CreER, Trp53^F, Braf^CA, and Rac1^LSL−P29S allele combination. (H) Melanoma-free survival curves of Tyr-CreER^+/–;Trp53^F/F;Braf^CA/WT;Rac1^WT/WT mice and Tyr-CreER^+/–;Trp53^F/F;Braf^CA/WT;Rac1^{LSL−P29S/WT} mice were compared using log rank testing (Mantel-Cox). (I) Effects of RAC1^P29S on tumor number in Tyr-CreER^+/–;Trp53^F/F;Braf^CA/WT mice. Groups compared using an unpaired two-tailed t test (n = 7–9 mice per genotype). (J) Representative photographs and histology of the dorsal skin of various mouse melanoma models. All mice are Tyr-CreER^+/–, with additional alleles indicated. Time since 4OHT treatment is indicated. For all graphs: bars represent means ± SD. See also Figure S6.

Figure 7

Effect of RAC1^P29S Drug Resistance and Mesenchymal Phenotype in BRAF-Driven Melanoma (A) GSEA plot of the EMT hallmark gene set from MSigDB, using RNA-seq data of tumor lysates from Tyr-CreER^+/–;Pten^F/WT;Braf^CA/WT;Rac1^{LSL−P29S/WT} mice versus tumor lysates from Tyr-CreER^+/–;Pten^F/WT;Braf^CA/WT;Rac1^WT/WT mice (n = 6 tumors from five to six animals per group). (B) Normalized mRNA read counts of mesenchymal markers in tumor lysates from Tyr-CreER^+/–;Pten^F/WT;Braf^CA/WT mice. For statistical comparison of groups, Wald test was applied in combination with Benjamini-Hochberg correction (n = 6 tumors from five to six animals per genotype). (C) Effect of expression of RAC1^P29S in melanoma on tumor architecture. Representative tumor sections shown stained with H&E or melanoma cell markers S100 and SOX10 (immunohistochemistry). (D) Tumors were induced in Tyr-CreER^+/–;Pten^F/WT;Braf^CA/WT mice by topical 4OHT. After tumors were established, animals were treated with PLX4720 incorporated in chow. Tumor growth curves were compared using two-way ANOVA, with the p value for the genotype effect indicated (n = 18–44 tumors from 6 to 10 animals per data point). Representative photos of tumors are included. (E) Tumors at the end of the experiment presented in (D) were harvested, lysed and used for immunoblotting with indicated antibodies. (F) Influence of RAC1^P29S/L mutation status on relapse-free survival in patients with melanoma. Data from the TCGA cutaneous melanoma cohort (n = 309 patients, 45% BRAF^mut and 55% BRAF^WT) were used. Groups were compared using log rank testing (Mantel-Cox), with the p value indicated. (G) After tumors were established by topical 4OHT, animals were treated with PLX4720 incorporated in chow and CCG-257081 by oral gavage. Tumor growth curves were compared using two-way ANOVA, with the p value for the genotype effect indicated (n = 10–15 tumors from 3 to 4 animals per data point). (H) GSEA was performed using the four melanoma differentiation state signatures developed by Tsoi et al. (2018), in combination with RNA-seq dataset of changes in ER-RAC1^P29S melanocyte-treated 4OHT (n = 3 independent experiments; left panel), RNA-seq data from melanocytes with endogenous RAC1^P29S versus RAC1^WT (n = 3 independent cultures per genotype; middle panel) and RNA-seq data of tumor lysates from Tyr-CreER^+/–;Pten^F/WT;Braf^CA/WT;Rac1^{LSL−P29S/WT} mice versus Tyr-CreER^+/–;Pten^F/WT;Braf^CA/WT;Rac1^WT/WT mice (n = 6 tumors from five to six animals per group; right panel); p values are indicated. For all graphs: means ± SD is shown. See also Figure S7.